At approximately 8:30 a.m. on the first Friday of every month, the Department of Labor's Bureau of Labor Statistics releases a snapshot showing what the U.S. labor market looked like in the previous month. The report usually runs more than 30 pages, and is filled with (generally obscure) statistics, tables, and charts. But the highlight of each month's report is a single figure that circulates quickly through the press, and is taken immediately to be an important indicator of the general health of the economy: the unemployment rate.
During the recent recession in particular, the unemployment rate has been at the center of our economic-policy debates. In October 2009, unemployment peaked at 10.1% — the highest rate since the early 1980s. Even as overall economic performance has improved in recent months, the unemployment rate has remained quite high, still hovering just below 10%. And since being out of work influences voters much more than, say, a modest uptick in economic growth, it seems clear that, at least as a political matter, the Great Recession cannot be declared "over" until the unemployment rate dips significantly.
Meanwhile, because unemployment statistics can be such powerful rhetorical weapons — used to make the case that economic recovery has begun, or that policies aimed to spur the economy have failed — politicians and analysts have been pulling rather carelessly from the many figures in the monthly employment reports, cherry-picking numbers to support their various arguments. We increasingly hear, for instance, about the "real" unemployment rate — a number that includes people who have given up searching for work or who are "underemployed," and that stands at nearly 17%.
But why should "real" unemployment be considered a more useful economic barometer than the standard unemployment rate? Just what do these figures mean? What do they really count, and how reliable are they? As the economy is sure to take center stage in the months leading up to this fall's elections — and as the unemployment situation will no doubt serve as a proxy for our broader economic fortunes — we would benefit from a better understanding of what the seemingly familiar statistics actually tell us.
The federal government has collected general labor statistics for more than a century. The practice dates back to 1870, when the decennial census began counting the number of "gainfully employed" Americans, listing them by industry. Until 1930, these were the only useful government measures of employment; these early figures, though, counted the Americans who were working, rather than those who were not. "Unemployment" was not yet in the American lexicon, and rough unemployment figures for that period can be determined only by subtracting the number of workers from the broader working-age population — hardly a way to extract reliable, meaningful data.
In response to labor-market volatility in the late 1920s, the 1930 census took the first real measure of unemployment (though both the concept and the methods used to assess it were still quite crude). Individuals were asked if they had been at work the previous day; men of working age who answered "no" were then considered to be unemployed.
The Great Depression fundamentally transformed the collection of labor statistics. Federal and state governments had enacted a great deal of legislation intended to aid unemployed workers; they obviously needed reliable ways of assessing whether the policies were effective, but data from the decennial census were not nearly timely enough to be of use. So statisticians, using newly emerging methods of surveying public opinion, developed questionnaires designed to count working-age people who had jobs, were out of work and looking for jobs, or were out of the labor market altogether, and thus not seeking employment.
In 1940, these questionnaires were merged into one national survey, called the Monthly Report of Unemployment. The new survey was conducted and published under the auspices of the Works Progress Administration, an agency created to implement and track many of the New Deal's relief programs.
The survey also changed the way the government defined unemployment. The vague "gainful employment" threshold was replaced with a more complex set of terms meant to pinpoint a person's precise labor activity (or lack thereof) in the course of a month. To be included in the "labor force," individuals had to be employed, or unemployed but actively looking for work. Employment was defined as work that either produced compensation or provided a direct benefit to the economy, so unpaid family workers (like stay-at-home parents) were not included in the labor force. And a person who reported that he had no job, but had also not looked for work in the course of the month, was no longer counted as unemployed. This "active work" definition is still the basis of modern assessments of unemployment, since the government wants to determine the number of individuals who are actively seeking work but cannot currently find a job.
In 1942, as the Works Progress Administration was terminated, the task of producing the monthly report was transferred to the Census Bureau. The report was renamed the Current Population Survey in 1948, and is carried out to this day; it remains the source of much of the government's data about the labor force, including the monthly unemployment rate and other key measures. While the Census Bureau still collects the data, the Bureau of Labor Statistics has been responsible for analyzing and publishing it since 1959.
The Current Population Survey (sometimes called the household survey) looks at a random sample of 60,000 households each month, and examines the status of all working-age civilians who are not in institutional settings (like prisons or nursing homes). The enormous size of the sample allows BLS researchers to correct for statistical distortions — the survey's assessment of the unemployment rate, for instance, has a margin of error of less than 0.2% — and offers a reliable picture of the larger population. In order to establish a standard point of measurement, the survey inquires into the working status of an individual for the week that contains the 12th day of the month, and is generally conducted the following week, to make sure respondents clearly remember whether they were working or looking for work during the relevant period.
The CPS is a self-reported survey in which respondents (70% of whom are reached by telephone, and the rest through personal visits) are asked about the employment status and other characteristics of people in their households. This can lead to some reporting errors, since the survey respondents can differ over time and have different attitudes about their work prospects; the sheer size of the sample, though, helps to limit such errors. The CPS is used primarily to determine the size and demographic characteristics of the labor force, the occupations and industries in which Americans work, and the unemployment rate.
A number of other labor statistics, which are also included in the monthly jobs report, are gathered from a separate monthly survey called the Current Employment Statistics survey (often referred to as the payroll survey). The CES polls employers rather than workers, asking them to report on the number of workers they employ, as well as on their employees' hours and earnings.
Government surveys of employers in certain key industries date back to 1915, but the comprehensive CES survey — which seeks to provide a reliable snapshot of the entire labor market, not just a few occupations — didn't begin until the mid-1930s. It uses a much larger sample than the CPS, consisting of around 140,000 private and public employers of various sizes, which account for about 410,000 individual work sites and almost a third of all non-farm workers. The information is gathered with the aid of the labor or employment departments of all 50 states, which receive and process unemployment-insurance forms from essentially every employer, and therefore possess enormous quantities of raw information. Like the CPS, the CES survey measures the number of paid workers during the week that includes the 12th day of each month.
In essence, the CES survey measures jobs, while the CPS measures workers. The CPS is therefore a more reliable source for unemployment statistics — since, for instance, a person with two jobs could be counted twice in the CES survey but only once in the CPS, and self-employed Americans are not counted in the CES measurement at all. Nevertheless, the CES survey provides a treasure trove of useful data about the nature of the labor market, and its enormous sample size makes the statistics it provides very reliable.
Together, these two surveys form the official government assessment of the labor market. They offer a wealth of information — but precisely because they offer so much raw data, they must somehow be repackaged into more digestible portions. This of course raises the question of exactly which portions should be taken as the most useful measures of unemployment — the answer to which carries enormous economic and political significance.
Because unemployment plays such an important role in shaping our perceptions of the economy, experts have long raised concerns about the way the government defines the term. The basic unemployment rate today measures the number of unemployed workers as a percentage of the total civilian labor force; the labor force in turn is defined as those who reported that they were employed in a given month's Current Population Survey, plus those who did not work in the week about which the CPS inquired (but who did actively search for a job at some point in the previous four weeks and could begin work if offered a job).
Some labor economists argue that this definition overstates unemployment, because its criteria for "actively searching for a job" are exceedingly loose. These criteria count, for instance, a single casual conversation with a friend asking if he knows of a job opening, or any time at all spent editing a résumé, as actively looking for work. No distinction is drawn between someone who visits dozens of potential employers and follows up with phone calls and someone who changes the font on a cover letter and never sends it anywhere.
Others, meanwhile, worry that the definition actually understates unemployment. The concern is that the criteria for being "unemployed" exclude people who might want to work but who have not been actively searching in the past four weeks — perhaps because they have no hope for employment at the time, or are unable to search because of, say, health problems or family commitments. The definition also excludes people who are underemployed — like those who want to work full time, but can find only a part-time job.
In an effort to respond to these concerns, the Bureau of Labor Statistics publishes a range of unemployment measures in its monthly reports. Since the monthly CPS and CES surveys make an immense wealth of data available, the BLS can slice and dice the numbers in different ways and let economists, journalists, policymakers, and the broader public use them as they wish.
Today, the agency publishes six different unemployment figures, which it terms U-1 through U-6; each offers a different measure of the under-utilization of labor-force resources. The so-called "U series" was originally created in 1976 by Julius Shiskin, who was then commissioner of the BLS, and was revised in 1994 to produce the categories still in use today.
The measures are numbered from the most restrictive definition of unemployment to the least. U-1 includes workers who have been unemployed and actively searching for jobs for 15 weeks or more; it is used mostly to assess the severity of unemployment, since significant long-term unemployment suggests a particularly depressed labor market. U-2 is a measure of job loss, counting the percentage of the labor force that is unemployed as a result of having been involuntarily let go from a job. (This figure, too, is often used as a measure of the severity of problems in the labor market.)
U-3 is the familiar official unemployment rate, measuring the number of people who are out of work but actively looking for jobs as a percentage of the total labor force. It is by far the most commonly used and cited labor statistic in America, and remains the essential benchmark for professionals and the general public alike. Measured and defined in the same way since the late 1940s, it also allows for helpful comparative assessments of the labor market over the past several decades.
The U-4 rate includes the unemployed workers counted in the official U-3 rate, plus so-called "discouraged workers" — people who are not in the labor force and not looking for jobs because they do not believe they can find work. To be counted in this category, a respondent to the Current Population Survey must report that he wants a job and is available to work, and that he has searched for a job in the prior year, but is not currently looking because he believes the search would be futile. Commonly cited reasons for such discouragement include an individual's sense that he does not have enough education or training to obtain a new job; that there are no jobs where he lives; or that potential employers would discriminate against him because of his age, sex, or race, or for other reasons.
The discouraged-worker category is notoriously subjective, and efforts to measure it have sparked a great deal of controversy among labor economists over the years. The current definition is the result of work by the National Commission on Employment and Unemployment Statistics, appointed by the Carter administration in 1978 to review the BLS's methods; the commission called for the narrowing of an earlier definition, which had not reflected whether respondents were available to work if offered jobs. A minority of the commission's members actually wanted a portion of this newly defined discouraged-workers category — namely, discouraged workers who had actively sought work in the previous six months — to be included in the official unemployment figure; this, of course, would have significantly increased the number of people considered unemployed. A majority rejected the proposal, and the measure of discouraged workers is now used mostly by economists to follow theoretical slack in the labor market, and so to get a sense of how many workers might join the labor force if economic conditions improved significantly. In this May's jobs report, the standard (U-3) unemployment rate was 9.7%, while the U-4 rate was 10.3%.
The final two "U series" measures are broader still, including people who say they want to work but who are not actively looking for work (or are not available to work) for reasons other than discouragement. The U-5 rate includes all workers counted under U-4, plus all other so-called "marginally attached workers" — basically anyone who claims to want a job. (A BLS economist once quipped that the "marginally attached" category is so broad, it includes every person who has ever thought about one day having a job.) In the May 2010 jobs report, this rate was 11.0%.
Finally, the U-6 rate includes everyone counted under U-5, plus individuals who are working part time but would like to work full time — the so-called "underemployed." (In May, this rate was 16.6%.) The "underemployment rate" has become an increasingly popular statistic in political debates about the economy, as it provides a startlingly high figure that can be cited as reliable evidence of trouble in the labor market. The problem, however, is that this number is startlingly high only in relation to the levels of unemployment that the official unemployment rate — the much more restrictive U-3 figure — has accustomed us to seeing. The chief utility of the U-6 rate for anyone but labor economists, then, is often just its shock value.
For economists, these last two definitions of unemployment can help provide some insight into labor-market movements. In particular, the spread between U-5 and U-6 can show how quickly businesses are returning to normalcy after a recession, because it offers a way to gauge changes in the number of hours worked as well as in the number of workers hired. An increase in U-6, meanwhile, can provide evidence that employers are shifting more workers to part-time schedules in response to declining economic conditions. But beyond these limited assessments, the significance of the U-5 and U-6 numbers is far from clear — and surely not as great as many commentators recklessly suggest.
The rationale for using the alternative measures of unemployment — especially those that include discouraged, marginally attached, and underemployed workers — is that they offer more complete or accurate measures of some important aspects of the labor market. For example, the inclusion of discouraged and marginally attached workers is said to capture the true amount of slack in the labor market better than the standard U-3 rate, because the individuals covered by these alternative measures could well pursue jobs if the economy were stronger. This was the reasoning that led the Bureau of Labor Statistics in the 1970s to start measuring discouraged workers in the first place; it is also the reason some economists and lay observers of the labor market refer to these figures and use them in formulating policy.
What is often overlooked, however, is that these alternative measures of unemployment are useful only if the people they include share some of the key characteristics of the people normally counted as unemployed — that is, if they are on a continuum of desire for work, so that greater demand for employees would in fact draw more and more of these "marginally attached" workers into the labor force.
It is of course inherently difficult to know if this is the case — since both the definitions involved, and the answers provided to government surveys about attitudes and expectations, are unavoidably subjective. But based on what we can surmise from the available data, the assumptions made by champions of the U-4, U-5, and U-6 rates are probably not justified. As it turns out, categories like "discouraged worker" and "marginally attached worker" — classifications used so eagerly by politicians and pundits when the standard unemployment rate simply won't make their point — are in fact not very useful measures of the health of the labor market.
The most thorough examination of this question was carried out by Monica Castillo, a BLS economist, in the mid-1990s. To determine whether individuals considered "discouraged" or "marginally attached" were in fact interested in working, Castillo broke down an immense amount of raw CPS data and tracked the behavior of individuals in these categories in a tightening labor market.
She examined their behavior in 1994 and 1995 — a period during which the standard unemployment rate fell from 6.6% (in January 1994) to 5.6% (in December 1995). The U-4 rate, which includes discouraged workers, fell from 7% to 5.9% during that period, while the very expansive U-6 rate declined from 11.8% to 10.4%. (Recall that the U-4 rate includes the U-3 rate within it, and that the U-6 rate includes both, as well as marginally attached workers and some part-time workers.) The question Castillo sought to address, then, was how much of the reduction in the more expansive rates is distinct from the decline seen in the official unemployment figure.
Castillo's research indicated that, despite very robust growth in employment for people who had clearly expressed interest in a job at the beginning of the period in question, most of those who were only "marginally attached" to the labor force in 1994 were still only marginally attached in 1995. By and large, they had not moved into the categories of actively seeking employment or actually being employed. And while 53.1% of the people who were unemployed under the standard definition at the beginning of 1994 had jobs at the end of 1995 (and another 19.4% were still actively looking for jobs), just slightly over one-quarter (27.4%) of the discouraged workers in January 1994 had jobs at the end of 1995. More than a third (36.2%) of the discouraged workers had completely dropped out of the labor force and no longer wanted jobs in the midst of a growing economy and improving labor market.
Castillo's research strongly suggests that workers officially considered unemployed are more likely to stay in the labor force and to work in an economic recovery than individuals who simply wish they had jobs. The two groups do not seem to be composed of people with similar levels of interest in work and similar attitudes toward job opportunities.
Another study — conducted in 2000 by Yolanda Kodrzycki, a vice president of the Federal Reserve Bank of Boston — found that marginally attached workers are far more likely to stop wanting jobs than to actually find work. Kodrzycki found that, on average, between 1994 and 2000, 42.4% of marginally attached workers in each month's CPS reported that they no longer wanted jobs just a month later. Only 12.8% were employed a month later, and just 24.8% were actively seeking work. In comparison, more than one-quarter of the officially unemployed workers in each month's CPS had obtained jobs a month later.
Kodrzycki argued that much of this disparity can be attributed to the demographic characteristics of marginally attached or discouraged workers: Generally speaking, they tend to be older and less educated than the general population. This limits their ability to find jobs even in a surging economy; at the same time, however, these demographic characteristics — especially age — make such people less likely to want to work in the first place.
These differences help explain why the alternative unemployment figures have tended to remain quite high even in periods when the official unemployment rate is very low. They also help explain what drives most of the changes in these alternative rates: namely, underlying changes in the standard U-3 unemployment rate — not shifts in the attitudes or circumstances of "discouraged" or "marginally attached" workers. The U-6 rate generally follows the movement of the U-3 rate, and although the spread between the two does tend to grow in serious downturns (as it has in this one), it does not do so in a way that suggests that underemployment offers a more meaningful or more "real" picture of the labor market than the standard unemployment rate.
THE DANGER OF ALTERNATIVE RATES
The problems with the alternative unemployment rates are of more than simply academic concern. In the recession of the past few years — and especially in recent months, as the labor market has begun at last to improve — almost every monthly jobs report has been followed by a torrent of stories, press releases, and statements insisting that the "real" unemployment rate is much higher than the standard number, or that "underemployment" is the real threat t o the economy.
Some reporters and pundits use the alternative measures of unemployment to make the labor-market situation seem even more desperate than it really is. "Don't be fooled by the fact that the official U.S. unemployment rate in February remained unchanged from January at 9.7 percent," the Washington Post's Frank Ahrens wrote in March. "The truer measure of unemployment in the United States rose from 16.5 percent in January to 16.8 percent in February."
Some conservatives have used similar arguments to buttress their criticisms of President Obama's economic policies. Also in February, Douglas Holtz-Eakin — a former director of the Congressional Budget Office, and an advisor to John McCain's 2008 presidential campaign — released a memo arguing that "the statistic to be closely watched is the underemployment number, which provides the truest measure of employment." Later that month, House minority whip Eric Cantor made the same point in a USA Today op-ed, writing that "while official unemployment hovers around 10%, real unemployment is frighteningly higher at nearly 17%."
Liberal politicians have done much the same, occasionally touting the alternative rates as evidence that the government needs to adopt more aggressive social policies to combat unemployment. These arguments have generally been muted over the past two years, since Democrats in Washington are not inclined to highlight grim jobs numbers on their own watch. But they could certainly be found in the left's rhetoric during the presidency of George W. Bush. A few days before the 2008 election, House majority leader Steny Hoyer told reporters that while "unemployment has been the focus of many news stories throughout the recent economic downturn," it was underemployment that really told the story of the Bush administration's economic failures — "which is why it is essential that Congress pass further economic recovery legislation to fight unemployment, underemployment, and ensure that those being hit hardest receive the help they need." Some prominent liberal economists — including Jared Bernstein, formerly of the Economic Policy Institute and now Vice President Joseph Biden's chief economic advisor — encouraged such views, arguing throughout the Bush years that the "real unemployment rate" called for more aggressive government spending to spur hiring.
This misinterpretation is not caused by inaccuracies in the broader measures of unemployment. The raw data from which they are calculated are highly reliable; moreover, the Bureau of Labor Statistics is very clear about what these numbers do and do not signify, and can hardly be blamed for their misuse. The problem is that their real value to the political debate is often vastly oversold. (The fact that people who do not want to work don't work, for instance, is hardly a great insight into the health of the economy.)
Champions of Keynesian fiscal policy should have learned in the course of the past two years that the broader unemployment measures are not their friends: They exaggerate the woes of the labor market, and make it difficult to portray any job-creation program as effective. And champions of limited government should see that, despite their temporary usefulness in scoring political points, these broader measures are surely not their friends either — for they seem to argue for ever more spending and ever more expansive welfare policies.
Most important, the broader unemployment rates are not helpful to analysts seeking to form the most accurate possible picture of the state of our labor market (and of the economy more generally). For that purpose — and for anyone who wants to engage in honest debate about the American work force — the standard unemployment rate, despite all its flaws, is very hard to beat.
Rea Hederman, Jr., is assistant director of the Heritage Foundation's Center for Data Analysis.
First appeared in National Affairs